5.1. Human Error Occurring during Manual Calculation
Wickens et al. [
25] divided human error into three types according to human information processing [
26].
Errors of recognition: Errors occurring during the information input by a sensory organ, including identification mistakes.
Errors of decision: Errors occurring during the decision-making of central nerves, including decision-making mistakes and memory failure.
Errors of performance: Errors occurring when one fails to complete an order from the motor center.
The aforementioned errors can occur not only in the process of manually constructing the geometry database but also in most office work. However, in the case of the geometry data, as the size of a building gets bigger, and its structure gets more complex, the amount of data increases exponentially. The extraction of the geometry data is an initial phase of the simulation process, so it is challenging for the simulation performer to recognize the occurrence of errors during the procedure, causing decreased credibility of the outcome.
Especially, the construction of the geometry database is the phase in which subjective judgments of the user are less likely to happen, so if a human error does not happen in the construction process, a certain degree of uniformity should be secured regardless of who the user is.
This study recruited three subjects and had each of them do both manual calculation and automated calculation through BIM modeling on three buildings (the module described
Section 3.2 paragraph was used). After this step, the degree of uniformity the two types of calculation methods represented was compared, and the causes of errors which lowers the degree of uniformity were examined.
5.2. Simulation Module, Recruitment of the Subjects and Experiment Method
The authors chose office buildings for which drawings were available and selected three of them with their sizes, as shown in
Table 1 and
Table 2 shows the images of modeling.
Model A is a two-story public building located in Daegu, and its total floor area is 551.84 m2. Model B is a mid-sized five-story office building located in Changwon, and its total floor area is 1696.95 m2. Model C is a five-story large-scale office building located in Gwangyang, and its total floor area is 6894.80 m2.
Three subjects were recruited to examine the degree of uniformity of the newly constructed databases from the three chosen models. The skill levels in CAD and BIM software of each subject were different from one another. Three individuals were recruited for the relative evaluation of the uniformity degree because a small group of subjects was believed to be better at thoroughly analyzing errors with an in-depth comparison than a large group of subjects.
Of the applicants working in the field of energy simulation for buildings, three individuals were chosen, and the skill level of each subject is shown in
Table 3. Subject A is an employee with experience of less than one year and is less skillful in CAD and 3D modeling than the other two subjects. Subject B has three years’ experience in the field and has an intermediate level of mastery in CAD and 3D modeling. Subject C has a career of over eight years in the same area and a high level of skills in both CAD and 3D modeling.
First, the three subjects are required to construct the geometry database on each model with manual calculations based on the drawings provided. Then, they compare the manually calculated database each subject constructed and calculate the uniformity of the three manual databases. In the next step, the three subjects create BIM models for automated calculation and calculate the uniformity of three automated databases (
Figure 6).
Table 4 illustrates the construction procedure of the geometry database of model 2 constructed by Subject B with automation calculation. The database comprises the space names, the architectural elements including walls, windows, floors, roofs and doors, and the surface, the interior/ exterior location, and the cardinal points of each element.
Each subject was given enough time for the construction of both a manual calculation database and an automated calculation database, two days for each calculation of each model, comprising 12 days in total.
5.3. The Calculation of the Uniformity and the Analysis of Error Causes
After the completion of the databases, the number of rooms and the number of architectural elements which Subject B calculated on each model are illustrated in
Table 5. The numbers of rooms of model A, model B and model C are 10, 38 and 63, respectively. The number of architectural elements increases with the number of rooms, 38, 145, and 221, respectively. This study considers that as the number of elements of each model increases, its complexity increases too, defining the complexities of model A, model B, and model C as low, middle, and high, respectively.
After the completion of the databases, the database each subject constructed was compared with those of the other two subjects to find if there was any inconsistency among the calculation outcomes of the element, the surface area. Such inconsistency was defined as an error.
The degree of uniformity was calculated by dividing the total number of elements into the number of elements having no errors, as shown in
Table 6. The uniformity values of model A, one is a manual database, and the other an automated one, agreed with each other, 97.94%. The two uniformity values of model B varied with 83.82% for a manual database and 99.20% for an automated one. In the case of model C, 87.48% for a manual database, and 98.94% for an automated one were obtained. Both model B and model C have higher uniformity for an automation database than a manual one.
The model with the middle- and high-complexity had the uniformity of less than 90% when manually calculated. On the other hand, the uniformity of an automation database stayed over 97% regardless of the complexity.
To determine the causes of the lower uniformity of manually calculated data, the close comparison was carried out studying the drawings, and the types of errors were categorized, and the numbers of errors were counted.
Table 7 shows the types and numbers of errors.
The errors were categorized into the human error type which was a mistake and the intended error which was caused by the subject’s subjective judgment. The human error type belongs to the error of recognition and the error of judgment, and the intended error includes the behavior where the subject arbitrarily decreased the number of rooms or altered the space boundary to minimize effort and time needed for the modeling task within limits where the behavior was judged not to influence the simulation outcome by the subject. In the manually calculated databases, human errors occurred due to error of data input, omission of data, calculation mistake and intended errors occurred due to disagreement of the number of rooms and disagreement of the space boundary. The automatically calculated database had human errors of missing element and intended errors of disagreement of the number of rooms.
The error of disagreement of the number of rooms occurred when the subject judged the thermal bridge effect would not occur between two rooms with the same usage and when two or more rooms were modeled as one. The error of disagreement of the space boundary occurred when space with high complexity and the space boundary were simplified for the ease of modeling.
The table above shows that the number of human errors occurring in the manually calculated database is 13.75 times that of the automation data and around 8.67 times more intended errors occurred in the manual calculation database than in the automatically calculated database. In addition, the total number of errors tends to increase rapidly at the point where the complexity of a model reached the middle level. That indicates that with the manual calculation, not only human errors occur in the database, but it is also easy for the intention of the user to intervene in a modeling activity.
Table 8 shows the number of errors each subject made. The number of intended errors increases with the level of skillfulness of the subject because they are caused by the subjective judgment of the user. The number of human errors, however, did not increase with the mastery level of the subject. The aforementioned cases have an unexpected impact on simulation, and manually calculated databases, which have high chances of human error occurrence, and may result in low credibility of the data and the simulation outcome.
The construction of the geometry database automatically extracted with high credibility also decreases process load in the error checking phase.
Figure 7 shows the automatic energy simulation process of data checking the user carries out when an error occurs.
On the premise that the geometry database has a certain level of credibility when an error occurs, the user does not have to carry out data checking on the geometry data. All the user has to do is check manually input variables such as material property, cooling, and heating system and work schedule. As the size of a building increases and the structure becomes more complex, the number of architectural elements which constitute the geometry data increases exponentially. Hence, the exemption of the geometry data from the data checking list is expected to decrease the time and cost needed in energy simulation.